127

MAGNETIC RESONANCE IMAGING CLINICS

Magn Reson Imaging Clin N Am 14 (2006) 127140

Imaging of Hemorrhagic Stroke

Eric E. Smith, MD, FRCPCa,b,*, Jonathan Rosand, Steven M. Greenberg, MD, PhDb,c
-

MD, MS

a,b

CT appearance of intracerebral hemorrhage MR appearance of intracerebral hemorrhage Effect of hematoma evolution on MR imaging appearance of intracerebral hemorrhage MR imaging pulse sequences and intracerebral hemorrhage appearance Causes of intracerebral hemorrhage Effect of location on cause of intracerebral hemorrhage Brain stem hemmorhage Cerebellar hemorrhage Intraventricular hemorrhage Multiple simultaneous intracranial hematoma

Specic causes of intracerebral hemorrhage Hypertensive hemorrhage Cerebral amyloid angiopathy Warfarin-related hemorrhage Vascular malformations Hemorrhagic transformation of brain infarction Brain tumors Ruptured saccular aneurysm Cerebral contusion Imaging evaluation for hemorrhage in the acute stroke setting Summary References

Hemorrhagic stroke accounts for approximately 15% of all stroke and is classied according to anatomic compartmentalization as intracerebral hemorrhage (ICH) (approximately two thirds) or subarachnoid hemorrhage (SAH) (approximately one third) [1]. This article discusses the use of CT and MR imaging for the differential diagnosis of ICH; for a detailed review of SAH, which is typically attributed to aneurysm rupture or severe trauma, the reader is referred elsewhere [2]. Topics to be discussed include the CT and MR imaging appearance

of ICH, the differential diagnosis of ICH by location, and the imaging evaluation of acute stroke with regard to hemorrhage.

CT appearance of intracerebral hemorrhage

The CT appearance of hemorrhage is determined by the degree of attenuation of the x-ray beam, which is proportional to the density of hemoglobin protein (relative to plasma concentration) within the hematoma.

This article was originally published in Neuroimaging Clinics of North America 2005;15(2):25972. This work was supported by National Institutes of Health Grant K23 NS046327. a Vascular and Critical Care Neurology, Massachusetts General Hospital, Boston, MA, USA b Harvard Medical School, Boston, MA, USA c Neurology Clinical Trials Unit, Massachusetts General Hospital, Boston, MA, USA * Corresponding author. Stroke Service, VBK 802, Massachusetts General Hospital, 55 Fruit Street, Boston, MA 02114. E-mail address: eesmith@partners.org (E.E. Smith).
1064-9689/06/$ see front matter 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.mric.2006.06.002

mri.theclinics.com

128

Smith et al

Immediately following vessel rupture, the hematoma consists of a collection of red blood cells, white blood cells, platelet clumps, and proteinrich serum that has a heterogeneous appearance on CT with attenuation in the range of 3060 Hounseld units (HU), depending on the degree of plasma extrusion [3]. In this hyperacute phase, hemorrhage may be difcult to distinguish from normal cortex because of similar attenuation. Over minutes to hours, a brin clot forms with an increase in attenuation to 6080 HU (Fig. 1A) [3]. Clot retraction and extrusion of serum can further increase attenuation to as high as 80100 HU in the center of the hematoma. The degree of attenuation may be reduced in patients with severe anemia [4], impaired clot formation due to coagulopathy, or volume averaging with adjacent tissue. Vasogenic edema evolves around the hematoma within hours and may continue to increase for up to 2 weeks after hemorrhage onset [5]. Over the following days, cells and protein are broken down and scavenged by macrophages, leading to slowly decreasing attenuation, with the greatest decrease at the periphery of the hematoma

and more gradual evolution toward the center (Fig. 1B and C) [6]. Within 4 to 9 days, the hematoma attenuation decreases to that of normal cortex, and within 2 to 3 weeks to that of normal white matter [3]. The CT recognition of subacute intracerebral hematoma can be challenging because the attenuation is similar to that of normal brain tissue, although mass effect may still be present. MR imaging can conrm subacute hematoma. As time goes on, attenuation continues to decrease to levels below that of the normal brain. Eventually, the hematoma resolves into a uid-lled or slit-like cavity that may be difcult to visualize on CT (Fig. 1D). Contrast enhancement is not present in the initial days following ICH but may develop at the periphery in weeks to months [7], sometimes leading to diagnostic confusion with brain tumor or abscess. A blood-uid level may be seen in medium to large ICH within the rst hours after onset; the dependent portion displays higher attenuation (Fig. 2) due to sedimentation of cellular elements [8]. This nding may be more common in ICH caused by anticoagulation [9], but it is not specic and has also been described in ICH due to hypertension, trauma, tumor, or arterial-venous malformation. The association with shorter time interval from ICH onset, and in some cases with anticoagulation, has led to speculation that incomplete clotting is required for blood-uid level formation.

MR appearance of intracerebral hemorrhage

The physics of MR imaging of hemorrhage is complex; multiple reviews have covered this topic in detail [10,11]. A brief explanation is warranted here, however, because an understanding of the signal characteristics of hemorrhage, as well as their

Fig. 1. CT appearance of hemorrhage. Serial CT scans of right thalamic hematoma. (A) Acute ICH in the right thalamus with mean attenuation 65 HU. (B) CT performed 8 days later than (A); the periphery of the hematoma is now isodense to the brain while the center of the hematoma has mean attenuation 45 HU. (C) CT performed 13 days later than (A) shows continued evolution of the hematoma with decreasing attenuation. (D) CT performed 5 months later than (A) shows a small area of encephalomalacia in the location of the previous hemorrhage.

Fig. 2. CT with blood-uid level. A 77-year-old woman was admitted with coma of 4 hours duration. CT scan shows massive left hemispheric hematoma with blood-uid level. No history of anticoagulation or coagulopathy.

Imaging of Hemorrhagic Stroke

129

evolution over time, is essential for radiologic interpretation. The MR signal intensity of hemorrhage is dependent on both the chemical state of the iron atoms within the hemoglobin molecule and the integrity of the red blood cell membrane [12]. Iron can be either diamagnetic or paramagnetic, depending on the state of its outer electron orbitals. In the paramagnetic state, it alters the T1 and T2 relaxation times of water protons through magnetic dipole dipole interactions and susceptibility effects. Dipoledipole interactions shorten both the T1 and T2 relaxation times but have a greater effect on T1. Susceptibility effect is present when iron atoms are compartmentalized within the red cell membrane,

causing magnetic eld inhomogeneity, with resulting loss of phase coherence and selective shortening of the T2 relaxation time. After degradation of red cell membranes, the iron becomes more homogenously distributed, and this effect is nullied. Other factors that inuence signal characteristics to a lesser extent include protein content, brain edema, oxygen tension, bloodbrain barrier breakdown, thrombus formation, and clot retraction [11]. Both the chemical environment surrounding the hemoglobin iron atom and red cell membrane integrity undergo relatively predictable changes after ICH. The following section enumerates these changes in MR signal characteristics during the different phases of ICH evolution (Fig. 3 and Table 1).
Fig. 3. MR imaging appearance of hemorrhage on T1-weighted (left column) and T2-weighted (right column) sequences for the different stages of hematoma (rows). Examples are selected from various patients. Hyperacute: There is relative isointensity on the T1-weighted and hyperintensity on the T2weighted sequence of this right occipital hematoma. A small degree of vasogenic edema surrounds the hematoma. On the T2-weighted sequence there is a thin rim of hypointensity that is barely detectable in the periphery; this is caused by susceptibility effect from deoxy-hemoglobin. Acute: The marked hypointensity on the T2-weighted sequence of this left frontal hematoma is caused by susceptibility effect from deoxy-hemoglobin. Early subacute: The hyperintensity on the T1-weighted sequence of this right occipital hematoma is caused by the oxidation of deoxy-hemoglobin to met-hemoglobin. Late subacute: The hyperintensity on the T2-weighted sequence of this large left frontal hematoma results from the loss of susceptibility effect caused by degradation of the red cell membranes. The degree of vasogenic edema is lesser compared with earlier phases. Chronic: A former large right frontal hematoma has resolved into a slit-like cavity with a rim of hypointensity on the T2-weighted sequence caused by hemosiderin deposition.

130

Smith et al

Table 1: Phase

Evolution of MR imaging signal characteristics with time Time Hours Hours to days Days to 1 week 1 week to months Rmonths Iron-containing molecule Oxyhemoglobin Deoxyhemoglobin Methemoglobin Methemoglobin Hemosiderin Iron oxidation state Fe21 Fe21 Fe31 Fe31 Fe31 Red cell membranes Intact Intact Intact Degraded Degraded T1 Y iso, Y [[ [[ iso, Y T2 [ Y Y [ Y Y T2* Y

Hyperacute Acute Early subacute Late subacute Chronic

Abbreviations: Fe, iron; iso, isointense relative to normal brain; [, hyperintense relative to brain, Y, hypointense relative to brain.

Effect of hematoma evolution on MR imaging appearance of intracerebral hemorrhage

Hyperacute phase The hyperacute phase of the hematoma is seen immediately following extravasation of blood into the brain parenchyma. At this stage the red cell membrane is intact, and the hemoglobin molecule is normally saturated with oxygen (oxy-hemoglobin). Specically, the iron atoms contained within the heme portions of the hemoglobin molecule are bound to oxygen. This is the only phase of hematoma in which the iron atoms have no unpaired electrons in their outer orbitals and are therefore diamagnetic, without exaggerated T1 relaxation or susceptibility effects. The ICH signal characteristics are thus not primarily attributable to iron but instead to the increased spin density of the hematoma relative to uninvolved brain tissue. Hyperacute hematoma appears slightly hypointense or iso-intense on T1-weighted images and slightly hyperintense on T2-weighted images (see Fig. 3); this pattern resembles that of many other pathologic conditions of the brain. Even early in the hyperacute phase, however, there is often deoxy-hemoglobin at the periphery of the hematoma, which appears as a thin rim of T2 hypointensity. This pattern can help differentiate hyperacute hematoma from other brain pathologies [1316]. Acute phase The acute phase, which begins within hours of ICH, is characterized by deoxy-hemoglobin. Deoxygenation occurs rst at the periphery of the hematoma and progresses toward the center. This pattern appears because intrahematoma oxygen tension is lowest in the periphery, where red cells are adjacent to oxygen-starved tissue, and highest in the center, because red cells do not use oxygen for their metabolism. The iron atoms of deoxy-hemoglobin have ve ligands and four unpaired electrons and hence are paramagnetic. Susceptibility effect is present

because the iron is compartmentalized within intact red cell membranes, resulting in hypointensity on T2-weighted images that is due to increased T2* relaxation (see Fig. 3). Magnetic dipoledipole interactions are prevented by the three-dimensional atomic structure of deoxy-hemoglobin, which blocks access of water protons to iron atoms. T1 relaxation times are therefore not shortened, and there is iso- or slight hypointensity on T1-weighted images (see Fig. 3). Sometimes a thin rim of T1 hyperintensity can be seen in the periphery of the hematoma, caused by early oxidation of intracellular deoxy-hemoglobin to intracellular methemoglobin. Early subacute phase After several days, the early subacute phase begins. The production of reducing substances declines with failure of red cell metabolism, and the iron atoms are oxidized to the ferric state, Fe31, to produce met-hemoglobin. Magnetic dipoledipole interactions can occur because the three-dimensional structure of met-hemoglobin exposes the iron atoms to water protons. This pattern leads to decreased T1 relaxation times and marked hyperintensity on T1-weighted images. Susceptibility effect is present because the red cell membranes remain intact, and hence there is continued hypointensity on T2-weighted images (see Fig. 3). Late subacute phase Over several days to weeks, the red cell membranes are degraded, and the late subacute phase begins. Susceptibility effect is lost because met-hemoglobin is no longer locally sequestered within red cell membranes; it freely diffuses within the hematoma cavity, resulting in a locally homogeneous magnetic eld. This pattern leads to T2* lengthening, and hence to increased hyperintensity, on T2-weighted images (see Fig. 3).

Imaging of Hemorrhagic Stroke

131

Chronic phase Over the ensuing months, the hematoma enters the chronic phase. The degree of hyperintensity on T1- and T2-weighted images lessens as the concentration of met-hemoglobin decreases with protein breakdown. The center of the hematoma may evolve into a uid-lled cavity with signal characteristics identical to cerebrospinal uid, or the walls of the cavity may collapse, leaving only a thin slit (see Fig. 3). As proteins are degraded, the iron atoms become liberated from the heme molecules, scavenged by macrophages, and converted into ferritin, which can be recycled. In most cases, however, the degree of iron deposition overwhelms the recycling capacity, with the excess being locally concentrated in hemosiderin molecules. The iron in hemosiderin does not have access to water protons and therefore exerts only susceptibility effect without signicant dipoledipole interactions, leading to marked hypointensity on T2-weighted images. This hypointensity is seen at the rim of the hematoma cavity and may persist indenitely. In practice, there is considerable variability in the orderly progression of hematoma signal change over time. The evolution of these signal characteristics may be inuenced by a number of factors, including ICH size, oxygen tension, integrity of the bloodbrain barrier, the presence of rebleeding, the efciency of the patients intrinsic repair processes, and the presence of an underlying lesion such as an arteriovenous malformation or tumor [11]. It is common to see different stages appear simultaneously. For these reasons, dating of bleed onset using MR imaging data alone is intrinsically imprecise.

MR imaging pulse sequences and intracerebral hemorrhage appearance

Hematoma signal characteristics are determined by the specic MR imaging pulse sequence applied. Higher magnetic eld strength increases sensitivity to susceptibility effects and therefore should allow

easier identication of hemorrhage. Fast spin-echo (FSE) sequences are less sensitive to magnetic susceptibility effects, owing to multiple 180 refocusing pulses, whereas echo planar imaging (EPI) and gradient recalled echo (GRE) sequences, which lack a 180 pulse, are more sensitive. Therefore, the use of FSE with relatively low-eld-strength magnets, a common situation in clinical practice, is associated with a lesser degree of T2-hypointensity than is that of EPI or GRE sequences on higherstrength magnets [10,17,18]. To overcome these limitations, GRE sequences should be used whenever the identication of hemorrhage is clinically important. The GRE sequence also known as susceptibility-weighted or T2*weighted sequenceemploys a partial ip angle without a refocusing pulse. In contrast to the use of the 180 spin-echo refocusing pulse, this method does not compensate for signal loss due to magnetic eld inhomogeneities, thus producing a stronger susceptibility effect. This pattern increases sensitivity for hematoma detection in the acute and chronic stages, because the already strong susceptibility effect causes extreme hypointensity on GRE sequences (Fig. 4) [10,19]. A relative disadvantage of the GRE sequence, however, is that artifactual signal loss is generated at the boundary of tissues that normally exhibit differences in susceptibility. This signal loss is particularly prominent at the pneumatized sinuses at the skull base and may obscure underlying lesions in those areas. Small hemosiderin deposits due to chronic, prior asymptomatic hemorrhage, sometimes referred to as microbleeds [20], are often only visualized on the GRE sequence. These may provide a clue to otherwise unsuspected underlying amyloid angiopathy, cavernous malformations, or hypertensive microvascular disease [2123]. Evidence suggests that, although most hemosiderin deposits persist indenitely, as many as 20% may become unapparent at 2 years [24]. Because most of these deposits are long lasting, the GRE sequence can be used to

Fig. 4. GRE sequence. T1(A) and T2-weighted (B) MR imaging sequences show acute hematoma of the right occipital lobe. Areas of hypointensity are seen on the T2-weighted sequence (B), caused by susceptibility effect from intracellular deoxy-hemoglobin. The susceptibility effect is far more conspicuous on the GRE sequence (C).

132

Smith et al

determine the cumulative hemorrhagic history of the patient over a prolonged period of time.

Effect of location on cause of intracerebral hemorrhage

The causes of ICH vary by location. The frequency of primary ICH at different sites, when not due to identiable structural lesions, is shown in Table 2. Supratentorial hemorrhages should be stratied into lobar and nonlobar ICH (Fig. 5). Lobar ICH is dened as hemorrhage at the supercial part of the brain, where bleeding is centered at the corticosubcortical boundary.Deep ICH is dened as hemorrhage in the deeper internal supratentorial compartment, with the hemorrhage centered principally in the putamen, head of the caudate nucleus, or thalamus. The site of origin can be difcult to determine when the hemorrhage is massive; in cases where both deep and subcortical structures are involved, the site of origin is more likely to be deep. A population-based study comparing lobar to nonlobar ICH showed that hypertension was a strong risk factor for nonlobar but not lobar ICH (relative risk 5 1.0) [26]. Conversely, apolipoprotein E genotype was a strong risk factor for lobar but not nonlobar ICH. The lack of association between hypertension and lobar ICH is notable and suggests that lobar and nonlobar hemorrhage have different causes. Most cases of deep ICH in the elderly are caused by hypertensive vasculopathy [27], whereas most lobar ICH in the elderly is caused by cerebral amyloid angiopathy (CAA). Pathologic evidence of CAA was found in 74% of lobar ICH patients over 55 years of age in a North American cohort [28]. Even among the elderly with hypertension, most lobar ICH was due to CAA [28]. In younger normotensive patients, particularly those aged less than 45 years [29], prevalence

Causes of intracerebral hemorrhage

Large cohort studies have identied the following risk factors for hemorrhagic stroke: age, hypertension, African-American or Hispanic ethnicity, smoking, excessive alcohol consumption, prior ischemic stroke, low serum cholesterol, and anticoagulant medications [25]. Age and hypertension account for the greatest risk to the population. It must be recognized, however, that ICH is not a single pathologic entity and may result from a number of diseases with differing pathophysiology and risk factors. It is the responsibility of the radiologist to recognize ndings that may support or refute the underlying differential diagnostic causes of ICH (Box 1). Radiologic clues to the cause of ICH may come from the topographic pattern, the signal characteristics, and the presence of other related lesions.
Box 1: Causes of intracerebral hemorrhage

Hypertension Cerebral amyloid angiopathy Vascular malformations Arteriovenous malformation Arteriovenous dural stula Cavernous hemangioma Hemorrhagic transformation of ischemic stroke Related to arterial infarction Related to venous infarction Vasculitis Moyamoya disease Coagulopathy Related to anticoagulant use Related to thrombolytic use Thrombocytopenia Decreased synthesis of clotting factors (eg, hemophilia, liver disease) Increased consumption of clotting factors (eg, disseminated intravascular coagulation) Brain tumor Aneurysm Ruptured berry aneurysm Ruptured mycotic aneurysm Related to sympathomimetic drug use Amphetamines Cocaine Phenylpropanolamine Ephedrine Trauma

Table 2: Location of primary of intracranial hematoma Location Lobar Striatum (putamen and caudate nucleus) Thalamus Cerebellum Multiple Brainstem Intraventricular No. of cases 114 89 64 34 14 12 3 % 35 27 19 10 4 4 1

Location of ICH in 330 consecutive patients age >18 y with primary ICH admitted to Massachusetts General Hospital between January 2001 and December 2004. Hemorrhages due to trauma, infarction, aneurysm, coagulopathy, tumor or vascular malformations were excluded. ICH in the striatum, thalamus, cerebellum and brainstem was most common (60%); these locations are typical for hypertensive hemorrhage.

Imaging of Hemorrhagic Stroke

133

Fig. 6. Brain stem ICH. Hemorrhage is present in the central pons with extension into the aqueduct of Silvius.

hematoma resection may be life saving. Causes of cerebellar ICH include hypertension, arteriovenous malformation, and, rarely, CAA [31,32].
Fig. 5. Lobar and deep hemispheric ICH. (A) Lobar ICH in the medial right occipital lobe. (B) Putaminal ICH originating from the posterior right putamen. (C) Left thalamic ICH dissecting medially into the ventricular system, with hemorrhage in the third and ipsilateral lateral ventricle. (D) Left caudate hemorrhage extending laterally into the white matter and medially into the ventricular system.

Intraventricular hemorrhage
Intraventricular hemorrhage is uncommon without parenchymal involvement. Before concluding an absence of parenchymal involvement, one should carefully examine the head of the caudate and thalamus, because even minute hemorrhage in these locations may quickly rupture into the ventricular system. Primary intraventricular hemorrhage has been associated with hypertension, anterior communicating artery aneurysm, anticoagulation, vascular malformation, moyamoya disease, and intraventricular neoplasm [33,34].

of both hypertensive vasculopathy and CAA is reduced. Therefore, other causes of deep and lobar ICH, such as vascular malformation, underlying tumor, underlying cavernous malformation, and hypertensive crisis induced by exogenous sympathomimetic drugs, should be considered in this age group.

Multiple simultaneous intracranial hematoma Brain stem hemmorhage

Brain stem hemorrhage (Fig. 6) most often occurs in the pons. Common causes include hypertensive vasculopathy, arteriovenous malformation, and cavernous malformation [30]. Mortality exceeds 60% when hypertension is the causative agent but is less for vascular malformation [30]. Multiple simultaneous ICH at different locations is uncommon and has been associated with

Cerebellar hemorrhage
Cerebellar hemorrhage (Fig. 7) is a neurosurgical emergency because the limited volume capacity of the posterior fossa leads to compression of critical brain stem structures. Typical complications of cerebellar hemorrhage include brain stem compression with cranial nerve palsy, respiratory arrest, upward and downward cerebellar herniation, and ventricular compression with acute obstructive hydrocephalus [31]. Prompt neurosurgical consultation is mandatory; suboccipital craniectomy with

Fig. 7. Cerebellar ICH. Hemorrhage is present in the left lateral cerebellum with mild surrounding vasogenic edema.

134

Smith et al

coagulopathy, infarction, tumor, CAA, vasculitis, and hypertension [35,36].

Specic causes of intracerebral hemorrhage

ICH due to hypertensive vasculopathy, CAA, or an unknown cause is often referred to as primary ICH to distinguish it from ICH due to other dened causes, such as vascular malformation, tumor, trauma, and infarction. The following section describes the imaging features of primary ICH, as well as several important causes of secondary ICH.

Hypertensive hemorrhage
Even in this era of improving blood pressure control, hypertensive vasculopathy remains the most common cause of ICH, although the incidence is declining [37]. Chronic hypertension causes a degenerative cerebral microangiopathy, characterized by hyalinization of the walls of small arteries and arterioles, and ultimately brinoid necrosis. Small areas of red cell extravasation may be associated with vessel wall cracking or microaneurysmal dilatation of the arteriole (also referred to as CharcotBrouchard aneurysms) [38]. ICH caused by hypertension most commonly results from rupture of the 50- to 200-mmdiameter lenticulostriate arteries that arise from the middle cerebral artery stem, leading to putaminal or caudate hemorrhage. It may also result from rupture of small perforating branches that arise from the basilar artery, leading to pontine or thalamic bleeds (see Figs. 5 and 6). Larger hematomas often dissect into the ventricles. Hypertensive vasculopathy is also a common cause of cerebellar ICH (see Fig. 7).

[47], and increased rate of new MR-detectable microbleeds [47]. CAA causes lobar hemorrhage in the cortex or subcortical white matter of the cerebrum [43] or, more rarely, the cerebellum [32]. Dissection into the subarachnoid space is common, whereas ventricular extension is uncommon. Rarely, CAA may present with solely sulcal SAH, thus mimicking aneurismal SAH [48]. CAA is frequently associated with clinically silent microbleeds remote from the symptomatic ICH (Fig. 8) [21]. Elderly patients with lobar ICH and multiple lobar microbleeds are highly likely to have CAA; the high specicity of this radiographic syndrome has been incorporated into a rating scale for diagnostic certainty of CAA [28]. The presence of any microbleeds in deep hemispheric locations should, conversely, put the diagnosis of CAA in doubt.

Warfarin-related hemorrhage
The strongest risk factor for warfarin-related hemorrhage is intensity of anticoagulation; age and history of ischemic stroke are additional independent risk factors [49]. The superimposed presence of leukoaraiosis appears to further increase risk [50,51]. A substantial proportion of elderly patients who have warfarin-related hemorrhage have underlying CAA [52]. Most studies have found no difference in hemorrhage location between patients taking and those not taking warfarin [53], although others have found an excess of cerebellar hemorrhage [54].

Cerebral amyloid angiopathy

CAA is caused by the deposition of b-amyloid in the arterial media/adventitia of the small arteries/arterioles in the meninges, cortex, and cerebellum. Affected vessels have eosinophilic walls that stain homogeneously with Congo red and demonstrate apple-green birefringence when viewed under polarized light [39]. The major risk factors for CAArelated hemorrhage are age and the presence of either the apolipoprotein E 34 allelewhich is associated with greater amyloid burden [40]or the apolipoprotein E 32 allele, which is associated with more severe vasculopathic change [41,42]. CAA-related hemorrhage is rare in persons aged less than 55 years, although the incidence increases exponentially in subsequent decades [43]. Recurrent hemorrhage is more frequent in CAA-related lobar ICH than in hypertensive ICH [44]. The risk of future recurrence is higher in patients with an apolipoprotein E 32 or 34 allele [45], moderate-to-severe white matter lesion burden [46], increased baseline number of MR imagingdetectable hemorrhages

Fig. 8. Clinically silent microbleeds in amyloid angiopathy. A 72-year-old woman presented with cognitive impairment; cortical biopsy showed amyloid angiopathy. MR imagingGRE sequence showed numerous cortical microbleeds with sparing of the deep hemispheric structures, including thalamus and basal ganglia. These lesions were not seen on the T2-weighted sequence. Sulcal vessels and calcication can also appear as small areas of hypointensity on the GRE sequence and must be distinguished from microbleeds.

Imaging of Hemorrhagic Stroke

135

Warfarin-related hematomas are more likely to expand [55] and to have fatal outcomes [56].

Vascular malformations
Vascular malformations that can bleed include arteriovenous malformations, arteriovenous dural stulas, and cavernous malformations (also known as cavernomas). Both venous angiomas (also known as developmental venous anomalies [DVAs]) and capillary telangiectasias are generally benign lesions, which are almost never associated with hemorrhage (although as many as 25% of DVAs are associated with an underlying cavernoma). Arteriovenous malformations and arteriovenous dural stulas can be difcult to detect without conventional catheter arteriography; suggestive but not sensitive imaging ndings include dilated feeding and draining vessels on MR T2weighted sequences, CT angiography, or MR angiography, as well as patchy enhancement (Fig. 9) [57]. Similarly, cavernous malformations are typically silent on all imaging modalities unless they have recently or previously bled. When recent, they may appear as popcorn-like lesions on T2weighted MR images because of the presence of multiple small hemorrhages of different ages arising from the same lesion (Fig. 10). When they are chronic, the hemosiderin from prior macro- or microhemorrhage may only be detectable on MR GRE sequences. These lesions may be multiple and familial [57].

produce mass effect and is usually asymptomatic, and parenchymal hematoma, which is more extensive and may be associated with neurologic deterioration [58]. Hemorrhage due to brain infarction may be recognized by the presence of surrounding cytotoxic edema conforming to an arterial territory, but it may be difcult to diagnose when early massive hemorrhage obscures the underlying infarct [59]. Venous infarction carries a higher risk for bleeding than arterial infarction, although anticoagulation treatment is typically indicatednot contraindicatedin the setting of venous thrombosis [60].

Brain tumors
Brain tumors are associated with neovascularity, incompetence of the bloodbrain barrier, and an increased risk for hemorrhage [61]. Tumors with a particular propensity to hemorrhage include glioblastoma multiforme, oligodendroglioma, and certain metastases such as melanoma, renal cell carcinoma, choriocarcinoma, and thyroid carcinoma (mnemonic: MR/CT). Lung cancer is also frequently hemorrhagic [6164]. In some cases, the tumor may be asymptomatic and unrecognized until presentation with hemorrhage. The CT and MR imaging characteristics of tumorassociated hematoma are often atypical and complex, because the blood may be of multiple ages and may be admixed with abnormal neoplastic tissue containing cysts and necrosis. Evolution of the MR signal changes is often delayed, possibly because of extremely low intratumoral partial pressure of oxygen, and hemosiderin formation may be absent [65,66]. The location may be atypical for hemorrhage caused by cerebrovascular disease, and there may be multiple simultaneous hemorrhages. The degree of vasogenic edema surrounding tumor-associated hemorrhage is greater than that in primary ICH and persists even into the chronic phase of hematoma [67]. Administration of contrast may reveal tumor enhancement (Fig. 11); the specicity of this enhancement for tumor is reduced in the subacute phase by enhancement of the hematoma capsule. In some cases, hemorrhage may completely obscure the underlying lesion; repeat imaging after hematoma resolution can allow tumor detection.

Hemorrhagic transformation of brain infarction

Infarcted brain tissue has a propensity to bleed. A common classication scheme differentiates between hemorrhagic infarction, which does not

Ruptured saccular aneurysm

Fig. 9. ICH due to arteriovenous malformation. CT angiogram (A) shows a large acute hemorrhage in the right hemisphere, with multiple feeding vessels in the periventricular white matter. There is a large draining vein along the lateral wall of the right lateral ventricle. Catheter angiography (B, anteroposterior view of right carotid injection) conrms the presence of arteriovenous malformation.

Blood from a ruptured saccular aneurysm enters the subarachnoid space under great pressure and may dissect into the brain parenchyma. Parenchymal hematoma is seen in 4% to 19% of patients with SAH due to saccular aneurysm and is highly correlated with the location of the ruptured aneurysm [68]. The most common locations are the medial frontal

136

Smith et al

Fig. 10. Cavernous malformation. T1- (A) and T2-weighted (B) MR sequences show a large cavernous malformation in the left frontal lobe. T2-weighted sequence (B) shows a heterogenous central core of variable hyperintensity surrounded by a deeply hypointense rim, caused by hemorrhage of various ages surrounded by hemosiderin-stained tissue. CT angiogram (C) reveals this lesion to be associated with a venous angioma.

lobe adjacent to a ruptured anterior communicating artery or anterior cerebral artery aneurysm (Fig. 12) and the temporal lobe adjacent to a ruptured middle cerebral artery aneurysm. In some cases, the amount of associated subarachnoid blood may be minimal. When ICH is immediately adjacent to the subarachnoid space at the base of the brain or basal interhemispheric ssure, vascular imaging should be strongly considered to exclude saccular aneurysm.

trauma, such as skull fracture, subdural hematoma, and epidural or subgaleal hematoma.

Imaging evaluation for hemorrhage in the acute stroke setting

Management decisions in the acute stroke setting rely on the differentiation of hemorrhagic from ischemic stroke. Imaging is required to make this differentiation, because there are no clinical features that reliably predict hemorrhagic stroke [1]. Demonstration of hemorrhagic absence identies ischemic stroke patients who may be eligible for thrombolysis, whereas demonstration of hemorrhagic presence may, in the future, identify acute ICH patients eligible for medical therapies to prevent continued bleeding. One such therapy currently under investigation is recombinant activated factor VII, which has shown promising results in a phase II study [70]. CT has traditionally been preferred over MR imaging for identication of ICH. CT scanning is faster, less expensive, and more widely available; it can be safely performed in patients with contraindications

Cerebral contusion
Brain contusion deserves mention because of the potential for misclassication as hemorrhagic stroke when a history of trauma cannot be elicitedfor example, in a patient found alone and confused. Contusions frequently occur in the basal anterior frontal and temporal lobes, where the brain is adjacent to the bony oor of the anterior and middle cranial fossa [69]. They may also be seen in the cortex either on the same side as the injury or as contrecoup. Contusions may be multiple and are often associated with other evidence of

Fig. 11. ICH originating within brain tumor. CT scan (A) shows acute hemorrhage with marked surrounding edema. T1-weighted MR imaging (B) shows hyperintensity within the hematoma, consistent with met-hemoglobin (subacute hemorrhage). Following contrast administration (C), there is a marked increased central hyperintensity on the T1-weighted sequence. Biopsy showed an underlying glioblastoma multiforme.

Imaging of Hemorrhagic Stroke

137

Fig. 12. ICH due to ruptured saccular aneurysm. CT scan (A) shows hemorrhage in the right medial basal frontal lobe. Subarachnoid hemorrhage is seen in both sylvian ssures and in the interhemispheric ssure. (A left skull defect with mild pneumocephalus is due to recent craniotomy for aneurysm clipping.) Maximum intensity projection images from the preoperative CT angiogram (B) reveal an anterior communicating artery aneurysm as the bleeding source.

to MR, including claustrophobia. Moreover, the sensitivity of conventional MR imaging for hyperacute hematoma detection has been questioned, because the oxy-hemoglobin stage of hemorrhage is isointense to water (see Table 1) [10]. Ample evidence indicates, however, that even in the earliest stages of hemorrhage, deoxy-hemoglobin is present in the lesion periphery, with corresponding hypointensity on T2 and GRE sequences [1316]. MR imaging has potential advantages over CT in the evaluation of ischemic stroke, both for delineating the extent of infarction with diffusion-weighted imaging (DWI) and detecting the presence of hemorrhagic complications with GRE sequences. MR may be superior to CT for the detection of early hemorrhagic conversion of infarction [7173], and, in patients who undergo intra-arterial thrombolysis, it may be better at distinguishing between hemorrhage and contrast extravasations (postlysis blush) into infarcted regions [74]. MR imaging is denitely more sensitive than CT for the detection of chronic microbleeds [71], which have been linked in at least one report to a higher risk of subsequent hemorrhagic transformation of infarction [75]. Although it has been suggested [76,77]though not proved [78]that the presence of baseline microbleeds could be a risk factor for major hemorrhage following thrombolysis, this hypothesis has yet to affect clinical management. Two blinded studies have compared CT and conventional MR imaging with MR GRE sequence for the detection of acute ICH in stroke patients. One study, which used acute ICH cases and selected ischemic stroke controls, found that MR imaging detected ICH with 100% sensitivity and 100% specicity [79]. The other study, which used a multicenter

prospective cohort design, found that GRE MR imaging was more sensitive than CT for both the diagnosis of hemorrhagic transformation and the detection of chronic microbleeds. As noted earlier, however, GRE MR imaging is sensitive but not specic: 3 of 29 acute ICH cases in this study were misclassied as chronic ICH. This lack of specicity may help to explain why inter-rater reliability for detection of ICH was lower for MR imaging (k 5 .75 to .82) than for CT (k 5 .87 to .94). Moreover, although GRE is sensitive for ICH detection, its sensitivity for SAH detection is dubious at best; the one case of SAH in the study discussed earlier was not detected by GRE [71]. The available data therefore suggest that it is feasible to use MR as the sole imaging modality for acute stroke evaluation, but that expert interpretation should be available and that caution should be exercised when excluding SAH. The relative role of CT angiography or MR angiography in the emergent evaluation of the ICH patient has not been fully dened. Vascular imaging has the potential to identify secondary causes of ICH that might require urgent surgical treatment, such as saccular aneurysm or vascular malformation [80]. In one study, extravasation of CT angiography contrast into the hematoma, possibly representing active bleeding, was seen in 46% of patients studied at a mean of 4.5 hours and was associated with increased mortality [81].

Summary
Neuroimaging by CT or MR is necessary for the detection of hemorrhagic stroke and provides important data regarding the cause of stroke. Serial changes in the CT and MR appearance of hematoma attributable to temporal evolution must be assessed to assure accurate diagnosis. Emerging evidence suggests that the use of MR imaging alone may be adequate for identifying hemorrhage in acute stroke patients and that GRE MR imaging is superior to both CT and conventional spin-echo MR imaging sequences for the detection of chronic microbleeds and hemorrhagic conversion of infarction.

References
[1] Mohr JP, Caplan LR, Melski JW, et al. The Harvard Cooperative Stroke Registry: a prospective registry. Neurology 1978;28(8):75462. [2] Lasner TM, Raps EC. Clinical evaluation and management of aneurysmal subarachnoid hemorrhage. Neuroimaging Clin N Am 1997;7(4): 66978. [3] Bergstrom M, Ericson K, Levander B, et al. Variation with time of the attenuation values of intracranial hematomas. J Comput Assist Tomogr 1977;1(1):5763.

138

Smith et al

[4] Kasdon DL, Scott RM, Adelman LS, et al. Cerebellar hemorrhage with decreased absorption values on computed tomography: a case report. Neuroradiology 1977;13(5):2656. [5] Inaji M, Tomita H, Tone O, et al. Chronological changes of perihematomal edema of human intracerebral hematoma. Acta Neurochir Suppl 2003;86:4458. [6] Dolinskas CA, Bilaniuk LT, Zimmerman RA, et al. Computed tomography of intracerebral hematomas. I. Transmission CT observations on hematoma resolution. AJR Am J Roentgenol 1977;129(4):6818. [7] Messina AV. Computed tomography: contrast enhancement in resolving intracerebral hemorrhage. AJR Am J Roentgenol 1976;127(6):10502. [8] Ichikawa K, Yanagihara C. Sedimentation level in acute intracerebral hematoma in a patient receiving anticoagulation therapy: an autopsy study. Neuroradiology 1998;40(6):3802. [9] Peger MJ, Hardee EP, Contant CF Jr, et al. Sensitivity and specicity of uid-blood levels for coagulopathy in acute intracerebral hematomas. AJNR Am J Neuroradiol 1994;15(2):21723. [10] Bradley WG Jr. MR appearance of hemorrhage in the brain. Radiology 1993;189(1):1526. [11] Atlas SW, Thulborn KR. Intracranial hemorrhage. In: Atlas SW, editor. Magnetic resonance imaging of the brain and spine. 3rd edition. Philadelphia: Lippincott Williams & Wilkins; 2002. p. 773832. [12] Gomori JM, Grossman RI, Goldberg HI, et al. Intracranial hematomas: imaging by high-eld MR. Radiology 1985;157(1):8793. [13] Linfante I, Llinas RH, Caplan LR, et al. MRI features of intracerebral hemorrhage within 2 hours from symptom onset. Stroke 1999;30(11): 22637. [14] Patel MR, Edelman RR, Warach S. Detection of hyperacute primary intraparenchymal hemorrhage by magnetic resonance imaging. Stroke 1996;27(12):23214. [15] Wiesmann M, Mayer TE, Yousry I, et al. Detection of hyperacute parenchymal hemorrhage of the brain using echo-planar T2*-weighted and diffusion-weighted MRI. Eur Radiol 2001;11(5): 84953. [16] Wintermark M, Maeder P, Reichhart M, et al. MR pattern of hyperacute cerebral hemorrhage. J Magn Reson Imaging 2002;15(6):7059. [17] Allkemper T, Tombach B, Schwindt W, et al. Acute and subacute intracerebral hemorrhages: comparison of MR imaging at 1.5 and 3.0 Tinitial experience. Radiology 2004;232(3):87481. [18] Wolansky LJ, Holodny AI, Sheth MP, et al. Double-shot magnetic resonance imaging of cerebral lesions: fast spin-echo versus echo planar sequences. J Neuroimaging 2000;10(3):1317. [19] Liang L, Korogi Y, Sugahara T, et al. Detection of intracranial hemorrhage with susceptibilityweighted MR sequences. AJNR Am J Neuroradiol 1999;20(8):152734.

[20] Fazekas F, Kleinert R, Roob G, et al. Histopathologic analysis of foci of signal loss on gradientecho T2*-weighted MR images in patients with spontaneous intracerebral hemorrhage: evidence of microangiopathy-related microbleeds. AJNR Am J Neuroradiol 1999;20(4):63742. [21] Greenberg SM, Finklestein SP, Schaefer PW. Petechial hemorrhages accompanying lobar hemorrhage: detection by gradient-echo MRI. Neurology 1996;46(6):17514. [22] Jeong SW, Jung KH, Chu K, et al. Clinical and radiologic differences between primary intracerebral hemorrhage with and without microbleeds on gradient-echo magnetic resonance images. Arch Neurol 2004;61(6):9059. [23] Kato H, Izumiyama M, Izumiyama K, et al. Silent cerebral microbleeds on T2*-weighted MRI: correlation with stroke subtype, stroke recurrence, and leukoaraiosis. Stroke 2002;33(6):153640. [24] Messori A, Polonara G, Mabiglia C, et al. Is haemosiderin visible indenitely on gradient-echo MRI following traumatic intracerebral haemorrhage? Neuroradiology 2003;45(12):8816. [25] Smith EE, Koroshetz WJ. Epidemiology of stroke. In: Furie KL, Kelly PJ, editors. Current clinical neurology: handbook of stroke prevention in clinical practice. Totowa (NJ): Humana Press; 2004. p. 117. [26] Woo D, Sauerbeck LR, Kissela BM, et al. Genetic and environmental risk factors for intracerebral hemorrhage: preliminary results of a population-based study. Stroke 2002;33(5): 11905. [27] Fisher CM. Pathological observations in hypertensive cerebral hemorrhage. J Neuropathol Exp Neurol 1971;30(3):53650. [28] Knudsen KA, Rosand J, Karluk D, et al. Clinical diagnosis of cerebral amyloid angiopathy: validation of the Boston criteria. Neurology 2001; 56(4):5379. [29] Zhu XL, Chan MS, Poon WS. Spontaneous intracranial hemorrhage: which patients need diagnostic cerebral angiography? A prospective study of 206 cases and review of the literature. Stroke 1997;28(7):14069. [30] Rabinstein AA, Tisch SH, McClelland RL, et al. Cause is the main predictor of outcome in patients with pontine hemorrhage. Cerebrovascular Diseases 2004;17(1):6671. [31] Ott KH, Kase CS, Ojemann RG, et al. Cerebellar hemorrhage: diagnosis and treatment. A review of 56 cases. Arch Neurol 1974;31(3):1607. [32] Itoh Y, Yamada M, Hayakawa M, et al. Cerebral amyloid angiopathy: a signicant cause of cerebellar as well as lobar cerebral hemorrhage in the elderly. J Neurol Sci 1993;116(2):13541. [33] Angelopoulos M, Gupta SR, Azat Kia B. Primary intraventricular hemorrhage in adults: clinical features, risk factors, and outcome. Surg Neurol 1995;44(5):4336. [34] Marti-Fabregas J, Piles S, Guardia E, et al. Spontaneous primary intraventricular hemorrhage:

Imaging of Hemorrhagic Stroke

139

[35]

[36]

[37]

[38] [39]

[40]

[41]

[42]

[43] [44]

[45]

[46]

[47]

[48]

[49]

[50]

clinical data, etiology and outcome. J Neurol 1999;246(4):28791. Maurino J, Saposnik G, Lepera S, et al. Multiple simultaneous intracerebral hemorrhages: clinical features and outcome. Arch Neurol 2001;58(4): 62932. Tucker WS, Bilbao JM, Klodawsky H. Cerebral amyloid angiopathy and multiple intracerebral hematomas. Neurosurgery 1980;7(6):6114. Furlan AJ, Whisnant JP, Elveback LR. The decreasing incidence of primary intracerebral hemorrhage: a population study. Ann Neurol 1979; 5(4):36773. Fisher CM. Cerebral miliary aneurysms in hypertension. Am J Pathol 1972;66(2):31330. Vinters HV, Gilbert JJ. Cerebral amyloid angiopathy: incidence and complications in the aging brain. II. The distribution of amyloid vascular changes. Stroke 1983;14(6):9248. Greenberg SM, Briggs ME, Hyman BT, et al. Apolipoprotein E epsilon 4 is associated with the presence and earlier onset of hemorrhage in cerebral amyloid angiopathy. Stroke 1996;27(8): 13337. Greenberg SM, Vonsattel JP, Segal AZ, et al. Association of apolipoprotein E epsilon2 and vasculopathy in cerebral amyloid angiopathy. Neurology 1998;50(4):9615. McCarron MO, Nicoll JA, Stewart J, et al. The apolipoprotein E epsilon2 allele and the pathological features in cerebral amyloid angiopathyrelated hemorrhage. J Neuropathol Exp Neurol 1999;58(7):7118. Vinters HV. Cerebral amyloid angiopathy. A critical review. Stroke 1987;18(2):31124. Bailey RD, Hart RG, Benavente O, et al. Recurrent brain hemorrhage is more frequent than ischemic stroke after intracranial hemorrhage. Neurology 2001;56(6):7737. ODonnell HC, Rosand J, Knudsen KA, et al. Apolipoprotein E genotype and the risk of recurrent lobar intracerebral hemorrhage. N Engl J Med 2000;342(4):2405. Smith EE, Gurol ME, Eng JA, et al. White matter lesions, cognition, and recurrent hemorrhage in lobar intracerebral hemorrhage. Neurology 2004;63(9):160612. Greenberg SM, Eng JA, Ning M, et al. Hemorrhage burden predicts recurrent intracerebral hemorrhage after lobar hemorrhage. Stroke 2004;35(6):141520. Ohshima T, Endo T, Nukui H, et al. Cerebral amyloid angiopathy as a cause of subarachnoid hemorrhage. Stroke 1990;21(3):4803. Hylek EM, Singer DE. Risk factors for intracranial hemorrhage in outpatients taking warfarin. Ann Intern Med 1994;120(11):897902. Smith EE, Rosand J, Knudsen KA, et al. Leukoaraiosis is associated with warfarin-related hemorrhage following ischemic stroke. Neurology 2002;59(2):1937.

[51] Gorter JW. Major bleeding during anticoagulation after cerebral ischemia: patterns and risk factors. Stroke Prevention in Reversible Ischemia Trial (SPIRIT). European Atrial Fibrillation Trial (EAFT) study groups. Neurology 1999;53(6): 131927. [52] Rosand J, Hylek EM, ODonnell HC, et al. Warfarin-associated hemorrhage and cerebral amyloid angiopathy: a genetic and pathologic study. Neurology 2000;55(7):94751. [53] Hart RG, Boop BS, Anderson DC. Oral anticoagulants and intracranial hemorrhage. Facts and hypotheses. Stroke 1995;26(8):14717. [54] Kase CS, Robinson RK, Stein RW, et al. Anticoagulant-related intracerebral hemorrhage. Neurology 1985;35(7):9438. [55] Flibotte JJ, Hagan N, ODonnell J, et al. Warfarin, hematoma expansion, and outcome of intracerebral hemorrhage. Neurology 2004; 63(6):105964. [56] Rosand J, Eckman MH, Knudsen KA, et al. The effect of warfarin and intensity of anticoagulation on outcome of intracerebral hemorrhage. Arch Intern Med 2004;164(8):8804. [57] Barnes B, Cawley CM, Barrow DL. Intracerebral hemorrhage secondary to vascular lesions. Neurosurg Clin N Am 2002;13(3):28997. [58] Fiorelli M, Bastianello S, von Kummer R, et al. Hemorrhagic transformation within 36 hours of a cerebral infarct: relationships with early clinical deterioration and 3-month outcome in the European Cooperative Acute Stroke Study I (ECASS I) cohort. Stroke 1999;30(11): 22804. [59] Smith EE, Fitzsimmons AL, Nogueira RG, et al. Spontaneous hyperacute postischemic hemorrhage leading to death. J Neuroimaging 2004; 14(4):3614. [60] de Bruijn SF, de Haan RJ, Stam J. Clinical features and prognostic factors of cerebral venous sinus thrombosis in a prospective series of 59 patients. For the Cerebral Venous Sinus Thrombosis Study Group. J Neurol Neurosurg Psychiatry 2001; 70(1):1058. [61] Little JR, Dial B, Belanger G, et al. Brain hemorrhage from intracranial tumor. Stroke 1979; 10(3):2838. [62] Kondziolka D, Bernstein M, Resch L, et al. Signicance of hemorrhage into brain tumors: clinicopathological study. J Neurosurg 1987; 67(6):8527. [63] Wakai S, Yamakawa K, Manaka S, et al. Spontaneous intracranial hemorrhage caused by brain tumor: its incidence and clinical signicance. Neurosurgery 1982;10(4):43744. [64] Mandybur TI. Intracranial hemorrhage caused by metastatic tumors. Neurology 1977;27(7): 6505. [65] Destian S, Sze G, Krol G, et al. MR imaging of hemorrhagic intracranial neoplasms. AJR Am J Roentgenol 1989;152(1):13744.

140

Smith et al

[66] Atlas SW, Grossman RI, Gomori JM, et al. Hemorrhagic intracranial malignant neoplasms: spin-echo MR imaging. Radiology 1987;164(1): 717. [67] Tung GA, Julius BD, Rogg JM. MRI of intracerebral hematoma: value of vasogenic edema ratio for predicting the cause. Neuroradiology 2003; 45(6):35762. [68] Abbed KM, Ogilvy CS. Intracerebral hematoma from aneurysm rupture. Neurosurg Focus 2003; 15(4):E4. [69] Adams JH, Doyle D, Graham DI, et al. The contusion index: a reappraisal in human and experimental non-missile head injury. Neuropathol Appl Neurobiol 1985;11(4):299308. [70] Mayer SA, Brun NC, Begtrup K, et al. Recombinant activated factor VII for acute intracerebral hemorrhage. N Engl J Med 2005;352(8):77785. [71] Kidwell CS, Chalela JA, Saver JL, et al. Comparison of MRI and CT for detection of acute intracerebral hemorrhage. JAMA 2004;292(15):182330. [72] Nighoghossian N, Hermier M, Berthezene Y, et al. Early diagnosis of hemorrhagic transformation: diffusion/perfusionweighted MRI versus CT scan. Cerebrovasc Dis 2001;11(3):1516. [73] Weingarten K, Filippi C, Zimmerman RD, et al. Detection of hemorrhage in acute cerebral infarction. Evaluation with spin-echo and gradientecho MRI. Clin Imaging 1994;18(1):4355. [74] Greer DM, Koroshetz WJ, Cullen S, et al. Magnetic resonance imaging improves detection of intracerebral hemorrhage over computed

[75]

[76]

[77]

[78]

[79]

[80]

[81]

tomography after intra-arterial thrombolysis. Stroke 2004;35(2):4915. Nighoghossian N, Hermier M, Adeleine P, et al. Old microbleeds are a potential risk factor for cerebral bleeding after ischemic stroke: a gradientecho T2*-weighted brain MRI study. Stroke 2002;33(3):73542. Chalela JA, Kang DW, Warach S. Multiple cerebral microbleeds: MRI marker of a diffuse hemorrhage-prone state. J Neuroimaging 2004; 14(1):547. Kidwell CS, Saver JL, Villablanca JP, et al. Magnetic resonance imaging detection of microbleeds before thrombolysis: an emerging application. Stroke 2002;33(1):958. Derex L, Nighoghossian N, Hermier M, et al. Thrombolysis for ischemic stroke in patients with old microbleeds on pretreatment MRI. Cerebrovasc Dis 2004;17(23):23841. Fiebach JB, Schellinger PD, Gass A, et al. Stroke magnetic resonance imaging is accurate in hyperacute intracerebral hemorrhage: a multicenter study on the validity of stroke imaging. Stroke 2004;35(2):5026. Eshwar Chandra N, Khandelwal N, Bapuraj JR, et al. Spontaneous intracranial hematomas: role of dynamic CT and angiography. Acta Neurol Scand 1998;98(3):17681. Becker KJ, Baxter AB, Bybee HM, et al. Extravasation of radiographic contrast is an independent predictor of death in primary intracerebral hemorrhage. Stroke 1999;30(10):202532.